Abstract

I will discuss how to combine generative modelling of chemicals with expert insight in drug design. I will start from improving modelling results based on feedback for proposed molecules from the human, often called human-in-the-loop machine learning or knowledge eliciation, and extend to asking what if the expert does not know. I will generalize to collaborative AI ‘sidekicks’ able to help other agents reach their goals, and finally to AI-assisted Virtual Laboratories for R&D.